High resolution ensemble analysis: linking correlations and spread to physical processes S. Dey, R. Plant, N. Roberts and S. Migliorini

A view of the 2nd and 3rd August COPE cases using storm-permitting ensemble MOGREPS-UK

S. Dey Supervisors: R. Plant, N. Roberts and S. MiglioriniCOPE science meeting29/04/2014

1Cases

0.1 0.5 1.0 2.0 4.0 8.0 16.0 [mm/hr]

2nd August 16Z3nd August 18ZNimrod radar derived rain rates

MOGREPS-UK2.2km resolutionConvection permitting12 membersDownscaled from MOGREPS-GMethods not specific to MOGREPS

Methods of ensemble characterizationSpatial Vertical correlations

Introduction to MOGPRES-UKStress that techniques could be applied to other ensembles

3Ensemble meanNot physically representative

Motivation123cv

Scale dependence Faster error growth at smaller scales (Hohenegger and Schr 2007, BAMS)Need ensembles at convective scale

Individual forecastsEnsemble meanPhysical meaningWhy spatial methods are neededProblems with physically interpreting ensemble mean4Method 1: spatial analysisOver what spatial scales are the forecasts acceptably similar?

?=LABVaries point by pointMeasure of spatial agreement

Apply to different variables and different vertical levels.Methodology (qualitative)

5Spatial comparisons

20 55 90 125 [ km]0.1 0.5 1.0 2.0 4.0 8.0 16.0 [mm/hr]

Ensemble scalesRadar rain ratesEnsemble-radar scales2nd August 16Z 3nd August 14Z 6Method1212121212121212123D variables (A,B)Horizontal divergenceHorizontal wind speedCloud fraction in each layerSpecific humidityTemperature

Surface Rain rates- Temporal correlations (C)

ABCzCorrelations over areaVertical correlations for 3D variablesChoose areaTemporal correlations with surface rain rates7Correlations: 2nd August, 14ZConvergenceDivergence-ve correlation +ve correlation

1 grid points (2.2km)13x13 grid points (28.6km)21x21 grid points (46.2km)

Div-rain correlations3 neighbourhood sizesNoise->Individual cells->one cell8Summary 0.1 0.25 0.5 1 2 4 8 16 32 >32

Mesoscale Convective System 27/07/2013Scattered convection 29/07/2013Bands of thunderstorms 23/07/20131

17th JulyOrganized thunderstorms (NOT COPE CASE)23rd JulyBands of thunderstorms (IOP 5)27th JulyMCS (IOP 7)29th JulyConvective showers (IOP 9)2nd AugustConvection along SW peninsula (IOP 10)3rd AugustConvection along SW peninsula (IOP 11)Alans slide 19Main questions

Can we develop scale appropriate, multivariate and physically meaningful methods for obtaining information from convection permitting ensembles?

Techniques useful for assessing model changesIncreased membership by time lagging? Including the UKV forecast in the ensemble?Global ensembleMore statistical approach

Other interesting investigations? Verification of spatial scales for different variablesComparison of correlation results with observations

Alans slide 210